ADDENDUM (EPA) - COMPARISON OF MODEL
ESTIMATES TO REAL-WORLD MONITORING DATA
Back to SVTC EcoMaps HomePage
Air toxics concentrations for 1990 were estimated in the Cumulative Exposure Project with the use of computer dispersion modeling, rather than by measurement. The best way to assess the accuracy of the modeled pollutant concentrations is to compare them with available real-world measurements.
To evaluate the modeled 1990 concentrations, EPA gathered more than 700 measurements of annual average air toxics concentrations for 1990, representing 19 air toxics in up to 81 locations each. At each monitor location, the modeled air toxics concentration was compared to the monitored concentration by calculating a model-monitor ratio (modeled concentration divided by monitored concentration). This analysis found that: 1) about two-thirds of the time, the modeled concentration is within a factor of three of the monitored concentration (i.e., the model-monitor ratio is between 0.33 and 3.0); 2) when the modeled concentration and monitored concentration differ by more than a factor of three, the model is almost always too low, rather than too high; and 3) the model underestimates monitored concentrations about three times as frequently as it overestimates. Several state air pollution agencies have reached similar conclusions in their own comparisons of the modeled concentrations to air toxics monitoring data. Some of these states do not have air toxics monitoring data for 1990, and therefore used measurements available for more recent years. Examples of these analyses and their findings include:
- Vermont: The State of Vermont found general agreement between modeled concentrations and concentrations measured in 1993 and 1994 for 10 air toxics in 5 locations. Out of the 50 comparisons, there were 12 cases of significant underestimation (model-monitor ratio less than 0.1), and one case of significant overestimation (ratio greater than 10).
- New York: The Northeast States for Coordinated Air Use Management (NESCAUM) analyzed monitoring data from the State of New York and found that modeled benzene concentrations were all within a factor of three of measured concentrations (model-monitor ratios between 0.33 and 3.0) at 10 monitoring sites. Six years of monitoring data (1990-95) were used in the analysis. The 10 monitoring sites are characterized as industrial, light industrial, urban, suburban, and rural. No clear tendency of underestimation or overestimation was observed.
- Mid-Atlantic States: An analysis prepared for the Mid-Atlantic Regional Air Management Association compared the modeled air toxics concentrations to 1996 measurements of up to eight air toxics at three monitors located in Philadelphia, New Brunswick, and Baltimore. Model-monitor ratios ranged from 0.66 to 3.0; sixteen out of the 21 comparisons had ratios of less than 2.0. Some variability in performance across different types of HAPs was noted.
It should be noted that these results represent comparisons for a limited number of air toxics in a limited number of geographical areas. While these results cannot necessarily be extrapolated to other air toxics in other areas, they provide the best indication available of how well the model results relate to real-world data. We recommend directly comparing ambient monitoring data, if available, when interpreting modeled outputs for a particular area.
Contact Daniel Axelrad at [email protected] for more information about EPA's CEP. Also, please visit their website (below).
Visit EPA's Cumulative Exposure Project website to learn more..
Website Created by
Michael R. Meuser, Clary-Meuser Research Associates